University of Minnesota Driven to Discover
U of MNUniversity of Minnesota
Center for Transportation Studies

Inexpensive Attitude-Determination Systems for UAV Applications

Presentation by Demoz Gebre-Egziabher, Dept. of Aerospace Engineering

October 12, 2004

Uninhabited aerial vehicles (UAVs) offer the potential to carry out numerous tasks for which piloted airplanes are unsuitable, either due to the high cost of conventional aircraft or dangers to human pilots. Presently, much work on UAV development is aimed at making the vehicles smaller and less expensive to build and operate. Demoz Gebre-Egziabher of the Department of Aeronautical Engineering is working on ways to build guidance components for this new generation of small, cheap flyers.

Accurate navigation, Gebre-Egziabher explained, requires two kinds of information: the location of the vehicle in three dimensions, and its orientation. Location can be accurately determined using the Global Positioning System (GPS). For an aerial vehicle, orientation can be described by three angles relative to a reference orientation (known as "Euler angles"). These are pitch (rotation perpendicular to the axis of motion), roll (rotation around the axis of motion), and yaw (rotation around the vertical axis).

In piloted aircraft, the traditional way of tracking orientation is to use mechanical gyroscopes; however, mechanical gyros are too large, heavy, and expensive for most UAV applications. Solid state rate gyroscopes are an alternative. These small, fast gyroscopes do away with the heavy rotating weight, so they are much smaller. But inexpensive solid-state gyroscopes produce various types of output errors, including random noise and "drift" in which they register a change in orientation when none actually occurs.

To compensate for the limitations of inexpensive solid-state gyroscopes, Gebre-Egziabher’s research uses an approach known as "complementary filtering" in which GPS data is used to reduce the amount of error generated by the gyroscopes. A GPS-based attitude determination system can be constructed by comparing locations of receivers on different parts of an aircraft; using high-accuracy GPS systems such as carrier-wave differential GPS, which can discriminate distances of a few millimeters, allows these antenna arrays to be located very close to one another. However, GPS systems cannot respond as quickly as gyro-based systems.

Because the gyroscopes are good at detecting sudden movements, a high-pass filter is used to capture this data while rejecting low-bandwidth data characteristic of drift errors. The GPS signal is used to periodically correct errors introduced into the measurements by the rate gyroscopes. The result is a robust system that is light, small, and inexpensive enough for UAV applications. This complementary filtering solution also allows the vehicle to "coast" through situations where the GPS receivers briefly lose the signal from orbiting satellites, using rate gyro information alone.

Gebre-Egziabher also described techniques that do away with the multiple antenna arrays necessary in a GPS-aided system like that described above, using a technique called "GPS-aided vector matching." The essence of this approach is computing solutions to the so-called "Wahba’s Problem": given two vectors whose components are known in two separate coordinate frames, it is possible to determine the relative attitude between the two frames. In this case, the vectors used are the Earth's magnetic field vector (measured by a magnetometer) and the gravitational vector (computed by subtracting local acceleration from the specific force vector, measured by an accelerometer). (To remove the effects of local acceleration, timing data from a GPS signal is used to calculate change in position over time.)

Gebre-Egziabher's seminar showed the importance of "fusion" between different types of sensor systems in serving the needs of advanced transportation applications. When the next generation of uninhabited aerial vehicles takes to the skies, they will rely on integrated approaches such as this to perform both critical and mundane tasks including traffic monitoring, wildlife and crop surveillance, and remote sensing.